1. Field of the Invention
This invention relates to techniques for inspecting reticles that are used in fabricating microelectronic devices through a microphotolithographic process. Particularly, the present invention relates to a method and an apparatus for detecting defects by emulating the operation of a specific photolithography tool in which a reticle is to be used. Such a tool may be an optical exposure system including optical steppers, scanners, and step-and-scan exposure systems. The invention is embodied in a method and an apparatus that can be used readily for inspecting reticles in the industrial environment.
More particularly, the present invention relates to a method and apparatus for inspecting phase shift masks used in a microphotolithographic process. A phase shift mask (PSM) uses destructive interference to control intensity of coherent light exposing a wafer during a manufacturing process. Closely spaced apertures are constructed in such a way that any light passing through any particular aperture will be 180 degrees out of phase with respect to light passing through adjacent apertures. As a result, any light that spills into the dark region between adjacent apertures will cancel out or destructively interfere. At the middle of the particular aperture, the light from the adjacent aperture is constructively interfering. The destructive interference reduces unwanted exposure in the center of the dark region and allows smaller features to be etched on wafers. This effect occurs also when the image is not focused, so that image degradation because of focus error is significantly reduced. In lithographic terms, it extends the depth of focus (DOF). The physics of PSM is described by M. D. Levinson, N. S. Viswanathan, and R. A. Simpson in Improving resolution in photolithography with a PSM, an IEEE publication, vol. ED-29, No. 12, December 1982.
The combined effect of the focus error and phase error can be simplified by the following description, graphically shown in FIG. 4. Focus error contributes to a global phase error δφ, and mask phase errors introduce Δφ to a programmed 180° phase region. At a region of constructive interference, the intensity is a result of an amplitude sum of a particular feature plus an adjacent feature with phase and focus contribution: A(180°+Δφ)+A(0°+180°+δφ). The phase mask error δφ will be positive for positive defocus, and assuming that the global phase error δφ approximates the mask phase error Δφ, the interference will be between corresponding phases (most constructive result). For the best focus the phase difference will be Δφ(a less constructive interference). For negative defocus, the phase difference will be 2·Δφ (the lowest amount of interference). The comparison of the positive defocus image to the negative one will give a much larger difference signal than a comparison of images taken with normal focus.
Regions that phase shift light may be created by etching or deposition techniques. According to one technique, the mask substrate may be etched to a precise depth such that the light passing through that region is phase shifted by 180 degrees. Phase altering materials also may be deposited on a substrate.
2. Description of the Related Art
Modem microelectronic devices are commonly produced using a photolithographic process. In this process, a semiconductor wafer is first coated with a layer of photoresist. This photoresist layer is then exposed to illuminating light using a photomask (for simplicity, the terms photomask, mask, and reticle will be used here interchangeably) and subsequently developed. After the development, non-exposed photoresist is removed, and the exposed photoresist produces the image of the mask on the wafer. Thereafter, the uppermost layer of the wafer is etched. Thereafter, the remaining photoresist is stripped. For multilayer wafers, the above procedure is then repeated to produce subsequent patterned layers.
Increasing the number of components in microelectronic circuits produced using the above photolithographic process requires the use of very high resolution images in photoresist exposure. The major limitations on the resolution of the image that can be projected on the photoresist are created by the illuminating light diffraction effects on the features of the mask and the limitations on the quality of the mask itself. The diffraction effects become important when the wavelength of the electromagnetic radiation used in the exposure of the photoresist becomes significant with respect to the size of the features of the mask being reproduced during the exposure. Increasing the resolution and decreasing the size of the reproducible features of the projected images may be achieved by decreasing the wavelength of the light that is being used in the photoresist exposure. For this reason, it is advantageous to use the electromagnetic radiation in the ultraviolet region of the spectrum, which corresponds to the shorter wavelength. In particular, ultraviolet i-line (365 nm), deep UV (248 nm), 193 nm, and 157 nm wavelengths have been used. Extreme Ultraviolet (11–13 nm) wavelengths are known; it is expected to be within contemplation that such wavelengths may be used as well.
Another method for increasing the resolution of the image is the use of RET (Resolution Enhancement Techniques) which include: off axis illumination, OPC (Optical Proximity Correction) reticles, and PSM (Phase Shift Mask) reticles.
It should be appreciated by those skilled in the art that to produce an operational microelectronic circuit, a mask must be as defect-free as possible, and preferably should be completely defect-free. Therefore, mask inspection tools are needed to detect various defects in the masks that can potentially reduce microelectronic circuit fabrication yields. Smaller feature sizes of the masks used in the microphotolithographic process, as well as the use of the phase shift masks and OPC masks, require more sophisticated tools for mask inspection. For instance, the inspection of phase shift masks requires not only finding “conventional” defects, such as particles, but also detecting errors in the thickness of various regions of the mask. Numerous systems for mask inspection have been developed in response to the growing demands of the electronic industry.
Early mask inspection tools used actual photoresist exposure to study the quality of the mask. According to this method, the mask is placed on the optical exposure system and used to actually expose the photoresist. The image obtained in this way then is studied to determine whether the mask performs to specifications. Because this method is expensive, time-consuming, and often inaccurate, it is uneconomical and inefficient.
Certain kinds of mask defects (called “surface” defects, for example, a particle on the surface of a mask) can be detected by inspecting the mask using the image of the mask produced by the light transmitted through the mask and the light reflected by one face of the mask. The mask inspection tool that uses this method acquires both images and analyzes them. The results of the analysis of the two images yield the information on the condition of the mask. Other systems use die-to-die comparison, die-to-database comparison, or reflected image to transmitted image comparison. In the die-to-die comparison method, the acquired images of a die of the mask are compared to the images of another die from the same mask. In the die-to-database method, the acquired images are compared to images that are simulated using the design specifications.
Such an inspection system can detect defects that may or may not print on the photoresist during the actual photolithographic process. The major drawback of this method is that it studies the physical structure of the mask independently of the optical image actually produced by the mask on the wafer. For instance, variations in the line width of the image that the mask produces frequently are higher than the corresponding variation in the line width of the mask itself. This phenomenon is called MEEF (Mask Error Enhancement Factor). Another example is PSM, in which there is no visible relation between phase error and the printed image. Therefore, it is desirable to relate the physical structure of the mask to the actual image that the mask creates on the photoresist, and to study directly the image that the mask actually produces.
In order to facilitate the evaluation of the mask performance during the mask development stage, IBM Corporation has developed a microscope called the Aerial Image Measurement System (AIMS™) that uses an aerial imaging method for mask evaluation. The Zeiss MSM100, a mask development tool, implementing AIMS™ technology, is available commercially from Carl Zeiss, GmbH of Germany. The MSM100 system can be used to evaluate the printability properties of newly developed masks.
An aerial imaging method is described in European Patent Application No. 0628806. According to this method, the inspection system simulates an optical exposure system that is used to expose the photoresist during semiconductor device fabrication. The optical system of the mask inspection device uses a set of the exposure conditions, used in the actual microphotolithographic process, to create an image that would be produced on the photoresist during actual device fabrication. In particular, the system matches the wavelength, the partial coherence of the exposure light, illumination aperture and the imaging numerical aperture NA of the optical exposure system. The created aerial image is magnified and detected using a CCD camera that is sensitive to ultraviolet radiation.
In addition to evaluation of the mask design, the use of the aerial imaging method permits the detection of the mask defects that would print during the actual microphotolithographic process. Almost any kind of defect on the reticle, including a particle on the transparent region, a pin-hole, a pin-dot, an edge truncation, etc., causes line width variation in the printed image. The term “line width” used herein describes a set of parameters of the image produced by the reticle on the photoresist, such as wire-to-wire distances, that determine whether the reticle is to be rejected as defective. The acquired aerial images are analyzed using the AIMS™ software, also developed by IBM. Despite all the above advantages, the Zeiss/IBM system has limited application as a printability review station for a set of detected defects by other inspection systems.
U.S. Pat. No. 5,481,624 describes a system that uses aerial imaging for die-to-database inspection of phase shift masks. According to the described inspection method, an aerial image produced by a phase-shift mask is verified against the original circuit pattern that was used in manufacturing the mask.
U.S. Pat. No. 5,795,688 discloses a system that uses an aerial imaging method for inspection of microphotolithographic masks having optical proximity corrections using a die-to-database comparison. In this system, an aerial image of a mask manufactured using the aforementioned optical proximity corrections is compared to an aerial image of the same mask obtained by simulation. Various defects in the mask, such as missing chrome, contamination, glass damage, phase defects, and transmission errors are identified as discrepancies between the two images. The simulation process takes into account optical proximity effects due to the limited resolution of the optical exposure system and the proximity effects due to the photoresist etching during the mask manufacturing process. The simulated aerial image can be obtained using the original mask design or, alternatively, using the mask design corrected for optical proximity effects.
Despite the above advances in the mask inspection technology, at the present time there is no inspection tool that would fulfill the demands of the industry. The IBM system is designed for mask development labs, and not for production stage mask inspection, and therefore does not possess adequate automation.
Also, the inspection methods based on die-to-database comparison that are used by the existing aerial imaging systems are not always effective, especially for highly complicated mask designs. The die-to-database comparison method uses models describing the behavior of an optical exposure system, as well as the effects of the etching used in the mask manufacturing process to produce the simulated image used in the mask inspection. However, the actual mask is different from the mask design because of limitations of the mask writing tool. As a result, there are limitations in the accuracy of the transformation from database to aerial image. Inadequate simulation can lead to a significant number of “nuisance” defects—the discrepancies between the acquired aerial image and the simulated image being caused not by the presence of actual defects in the mask, but by inadequacies in the simulation model. Nuisance defects can greatly complicate the mask inspection. For all the foregoing reasons, the limitations on the quality of the simulated images limit the performance of the aerial imaging inspection techniques that use the die-to-database comparison.
Accordingly, there is a need for a mask inspection system that would make it possible to detect errors in the line width of the image that the mask would actually produce on the photoresist.
The system also must be capable of detecting the phase defects and the presence of surface defects such as particles, contaminations, coating defects, and the like.
It also is desirable for the mask inspection system to provide speedy and reliable identification of the above mask defects. Such a system would be able to work efficiently in a clean manufacturing environment such as fabs and mask shops and increase the productivity thereof.